← All workshopsRead What Happens When Requests Arrive Together?

Model the System Before Building It

Find the worst queue during the upload burst

Start quietly before a late burst

Even a service with enough average capacity can accumulate unfinished image requests during a burst or worker outage.

BOTIE

Why cannot the unused capacity in the first interval cancel the later burst?
The inputs we are comparing

Requests arriving per interval

Short burst

  1. 02
  2. 17
  3. 21

Quiet start cannot store capacity

  1. 00
  2. 18

Requests that can finish per interval

Short burst

  1. 04
  2. 14
  3. 24

Quiet start cannot store capacity

  1. 05
  2. 13
Result for short burst
3
Choose an explanation to test
A temporary stoppage leaves work for later

This check uses the pieces from “Retain unfinished requests across intervals”.

BOTIE

Botie receives jobs over four equal time slots. Available service capacity falls to zero during the second slot. Starting with an empty queue, add each slot's arrivals and then serve up to that slot's capacity. Return the largest remaining backlog after any slot.
Requests arriving per interval
  1. 03
  2. 14
  3. 24
  4. 30
Requests that can finish per interval
  1. 05
  2. 10
  3. 23
  4. 38

Use a number, text, or an array matching the requested result. For a matrix, put each row in brackets. Saving your prediction runs your current construction on these inputs and checks the earlier examples too.

Build a construction in the workspace above first. This case will test those pieces.

Why could a trace finish empty while still exceeding the available queue space earlier?

Keep the full workload trace and its maximum; final backlog and average rate cannot certify burst handling.

This counts fixed-size requests in discrete intervals. It does not measure individual latency or implement queue limits and retries.

Work through the Python challenges →
Continue the courseWill the Model Fit on the Device? →
← Where Does the Prediction Pipeline Run Out of Room?